Pinned Loading

1. LLM-Tuning-Safety/LLMs-Finetuning-Safety LLM-Tuning-Safety/LLMs-Finetuning-Safety Public

   We jailbreak GPT-3.5 Turbo’s safety guardrails by fine-tuning it on only 10 adversarially designed examples, at a cost of less than $0.20 via OpenAI’s APIs.

   Python 240 29

2. CHATS-lab/persuasive_jailbreaker CHATS-lab/persuasive_jailbreaker Public

   Persuasive Jailbreaker: we can persuade LLMs to jailbreak them!

   HTML 258 20

3. reds-lab/Narcissus reds-lab/Narcissus Public

   The official implementation of the CCS'23 paper, Narcissus clean-label backdoor attack -- only takes THREE images to poison a face recognition dataset in a clean-label way and achieves a 99.89% att…

   Python 105 12

4. I-BAU I-BAU Public

   Official Implementation of ICLR 2022 paper, ``Adversarial Unlearning of Backdoors via Implicit Hypergradient''

   Jupyter Notebook 50 13

5. frequency-backdoor frequency-backdoor Public

   ICCV 2021, We find most existing triggers of backdoor attacks in deep learning contain severe artifacts in the frequency domain. This Repo. explores how we can use these artifacts to develop strong…

   Jupyter Notebook 41 6

6. reds-lab/Meta-Sift reds-lab/Meta-Sift Public

   The official implementation of USENIX Security'23 paper "Meta-Sift" -- Ten minutes or less to find a 1000-size or larger clean subset on poisoned dataset.

   Python 18 4